From f66329e35db56e87ed3d20c4913cabecfe6e4e7c Mon Sep 17 00:00:00 2001 From: ZehuaCao <47251317+Romanticoseu@users.noreply.github.com> Date: Thu, 14 Mar 2024 10:14:13 +0800 Subject: [PATCH] Fix multiple get_enable_ipex function error (#10400) * fix multiple get_enable_ipex function error * remove get_enable_ipex_low_bit function --- python/llm/src/bigdl/llm/transformers/convert.py | 8 -------- python/llm/src/bigdl/llm/transformers/loader.py | 2 +- 2 files changed, 1 insertion(+), 9 deletions(-) diff --git a/python/llm/src/bigdl/llm/transformers/convert.py b/python/llm/src/bigdl/llm/transformers/convert.py index cc69eacd..ad672926 100644 --- a/python/llm/src/bigdl/llm/transformers/convert.py +++ b/python/llm/src/bigdl/llm/transformers/convert.py @@ -634,14 +634,6 @@ def _optimize_pre(model): return model -def get_enable_ipex(low_bit): - _enable_ipex = os.getenv("BIGDL_OPT_IPEX") - _enable_ipex = (_enable_ipex is not None) and (_enable_ipex.lower() == "true") - qtype = ggml_tensor_qtype[low_bit] - _enable_ipex = _enable_ipex and (qtype == ggml_tensor_qtype["bf16"]) - return _enable_ipex - - def ggml_convert_low_bit(model, qtype, optimize_model=True, convert_shape_only=False, device="cpu", modules_to_not_convert=None, cpu_embedding=False, diff --git a/python/llm/src/bigdl/llm/transformers/loader.py b/python/llm/src/bigdl/llm/transformers/loader.py index 95d2ee76..4c3250f8 100644 --- a/python/llm/src/bigdl/llm/transformers/loader.py +++ b/python/llm/src/bigdl/llm/transformers/loader.py @@ -65,7 +65,7 @@ def load_model( # Load tokenizer tokenizer = tokenizer_cls.from_pretrained(model_path, trust_remote_code=True) model = model_cls.from_pretrained(model_path, **model_kwargs) - if not get_enable_ipex(low_bit): + if not get_enable_ipex(): model = model.eval() if device == "xpu":